When Not to Ask: A Defense of Choice-Masking Nudges in Medical Research

Abstract

In this article, we examine the legality and ethics of a controversial but widespread practice in clinical research: choice-masking nudges. A choice-masking nudge (CMN) exists when a research team explicitly obscures a meaningful choice from participants by presenting a default decision as the standard way forward. Even though an easy-to-use opt-out mechanism is available for participants who independently express concerns with the standard default, the fact that a default has been pre-selected is not made obvious to research participants. To opt out of the nudge, a participant must overtly request non-standard treatment. We argue that use of such nudges in medical research can be justified by their individual, collective, and social benefits, provided that they respect autonomy and satisfy our four further acceptability conditions. The structure of this Article is as follows. In Part II, we describe three controversial cases of CMNs in medical research. In Part III, we provide background on nudging and explain how our proposed CMNs fit into the existing literature on nudging and libertarian paternalism. In Part IV, we explain how the reasonable person standard as employed by United States research regulations can be used to support CMNs. In Part IV, we anticipate some of the strongest objections to CMNs by explaining how CMNs are compatible with a wide range of plausible accounts of autonomy. Finally, in Part VI, we discuss four additional core considerations an acceptable CMN must meet: legitimate policy goals; benefits outweighing harms; burdens distributed fairly; and absence of ethically superior feasible alternatives. We also analyze the three existing controversies explored in Part II and show how each would benefit from the conceptual clarity offered by our analytic framework. Medical research is complicated and can be difficult for participants to understand; thoughtfully designed CMNs can play an important role in gently guiding large numbers of research participants toward decision outcomes that really are best for them and their communities.

I. Introduction

This paper focuses specifically on nudging participants in medical research. Nudges exist in a wide variety—they come from our smartphones and our employers,¹ regarding things as different as food choice² and energy consumption³—but regardless of the source, all nudges attempt to influence nudgees’ choices by changing the way options are presented.⁴ Although nudging is all around us and even permeates medical care,⁵ its use in medical research prompts unique ethical and regulatory concerns. For instance, beyond the way researchers present information about costs and benefits, may they go so far as to implement default choices that participants would have to actively opt-out of to avoid? More concretely, may we nudge participants towards receiving medically actionable information they might not ask for? May we nudge new parents towards contributing their newborn infant’s blood spot for future research use? May we nudge participants to deposit their research data in repositories where other investigators can access them to help answer a broad range of secondary scientific questions? We argue that as long as certain acceptability considerations are satisfied, nudges that mask choices regarding those questions can be ethically and legally acceptable.

A choice-masking nudge (CMN) is a particular kind of nudge that exists in medical research when a research team explicitly obscures a meaningful choice from participants by presenting a default decision as the standard way forward. Even though an easy-to-use opt-out mechanism is available for participants who independently express concerns with the standard default, the fact that a default has been pre-selected is not made obvious to research participants. Consider, for example, a consent form stating that secondary genetic findings may be returned to research participants (and explaining what that means) without providing the participant with an opportunity to explicitly consent.⁶ In that case, the form does not advertise the fact that receiving secondary findings has been deliberately pre-selected. To opt out of the nudge, a participant would need to ask the researchers to receive non-standard treatment—in this case, to not receive any secondary genetic findings. Since all CMNs must give nudgees a fair chance of opting out, nudgees should be made aware (at least in general terms) that some options may have been pre-selected as defaults and that they may choose to select a different option. Consent forms could include a disclaimer suggesting that participants who think they might prefer a different treatment than the one indicated should raise any concerns with the research team (for example, “please ask a member of the research team if you have any questions or concerns about automatic return of secondary findings”). This kind of general disclaimer is not expected to significantly reduce the efficacy of the CMNs; on the contrary, research in other contexts has shown that informing individuals that they are being nudged does not tend to reduce the nudge’s effectiveness.⁷ [include transition into next paragraph].

There are four elements of CMNs that are worth emphasizing. First, Researchers are prohibited from masking information regarding the default choice.⁸ Researchers avoid masking such information when the issue in question and the default action are clearly described, but the choice regarding the default action is masked by not presenting an opportunity to make an explicit choice.⁹ Second, because CMNs are not binding, participation in research does not depend on agreeing to the default action.¹⁰ Third, the CMN must be easy to resist.¹¹ That means that the opt-out mechanism to pursue an alternative to the default action must be easy to use after the participant self-identifies a concern.¹² Fourth, our focus in this paper is on meaningful choices—either because they are likely to have non-trivial impact, or because a sizable number of participants are expected to have strong preferences about them (in contrast to the mundane choices that are routinely hidden from research participants, such as the method of contact with the research team). Ideally, CMNs will have the greatest effect on research participants who do not have strong prior preferences about the choice that is masked, since those who do have strong preferences are more likely to notice and opt out of the nudge.¹³ While we offer a variety of relevant considerations for analyzing the ethical acceptability of a CMN within a broad, flexible framework, this paper does not provide an algorithmic account of how these various considerations weigh against each other in every circumstance.

The following examples of CMNs are all controversial and likely to meet serious opposition, stemming from a few ethical concerns. First, the perceived primacy of autonomy in the bioethics literature may lead critics to claim that our analysis of CMNs gives short shrift to that principle, particularly regarding the autonomy interests of individuals who would prefer not to be nudged. Bioethicists commonly identify four primary moral principles: autonomy, nonmaleficence, beneficence, and justice.¹⁴ Autonomy, however, is often treated as first among equals, and any perceived infringement of a research participant’s autonomy is taken very seriously.¹⁵ In this paper, we push back against the idea that autonomy should be the foremost ethical concern when analyzing the appropriateness of CMNs. While autonomy is relevant and constrains the scope of acceptable CMNs, considerations of individual, collective, and societal benefit are also crucial to our ethical analyses (and may not always align with maximally promoting autonomy).

Second is the related issue of dealing with the tail of the curve. In most research cohorts, there will be a small group of participants who have outlier views or preferences.¹⁶ Respecting these differences to help this small group make more informed decisions could impact the larger research cohort, or even the research enterprise as a whole. In cases where nudges cannot be feasibly tailored to individuals, heterogeneity among participants makes it unlikely that all nudgees will benefit from or even be equally affected by the nudge.

Third, regulatory opposition will likely come from the perception that CMNs cut against the large body of jurisprudence that emphasizes informed consent in medical research. However, while the Federal Policy for the Protection of Human Subjects (the revised Common Rule) explicitly mandates disclosure of relevant information and an opportunity to discuss it, it does not mention active solicitation of preferences.¹⁷ Instead, our justification for CMNs invokes the reasonable person standard, which has often been interpreted to require preference solicitation.¹⁸

In this paper, we draw a distinction between nudges with implications for information use (informational nudges) and nudges with direct physical implications (physical nudges). The former concerns actions regarding the treatment of information gained from research, while the latter have to do directly with physical procedures that have significant implications for bodily integrity.¹⁹ Thus, a nudge concerning drawing a blood sample for clinical care would be physical, while a nudge concerning what is to be done with what remains of the leftover sample would be informational. We claim that it is reasonable to not explicitly solicit consent for aspects of research with predominantly informational implications, provided that there is sufficient information disclosure, that our other acceptability considerations are met, and that doing so is consistent with current US research regulations.

The remained of this paper is organized as follows. In Part II, we present three cases of medical research decisions where CMNs could be implemented. In Part III, we provide background on nudging and explain how our proposed choice-masking nudges fit into the large existing literature on nudging and libertarian paternalism.²⁰ In Part IV, we explain how the reasonable person standard as employed United States research regulations can be used to support CMNs.²¹ In Part IV, we anticipate some of the strongest objections to CMNs by discussing prominent features of various views of autonomy, and explaining how CMNs are compatible with a wide range of plausible accounts of autonomy.²² Finally, in Part VI, we lay out four core considerations relevant to the acceptability of CMNs: legitimate policy goals; benefits outweighing harms; burdens distributed fairly; and absence of ethically superior feasible alternatives.²³ We also analyze the three cases presented in Part II, in light of those considerations. A CMN that respects autonomy and that satisfies the considerations relevant to acceptability, we argue, can plausibly be justified on the basis of individual, collective, or social benefit.²⁴

Part II: Three Cases of Choice-Masking Nudges in Research

A. The Right Not to Know Genetic Information About Oneself

In the past fifteen years, advances in genomic sequencing technology have given us the capacity to rapidly and inexpensively sequence a person’s entire genome.²⁵ Physicians and researchers quickly began incorporating this powerful tool into their arsenal.²⁶ While next-generation sequencing has radically changed how science is conducted, researchers and Institutional Review Boards (IRBs) have struggled with the ethical implications of this new technology.²⁷ In particular, given that massive amounts of data generated by genomic sequencing can contain clinically actionable findings unrelated to the specific condition being investigated, the research community has vigorously debated about whether and how to honor participants’ so-called right not to know (RNTK) genetic information about themselves.²⁸

For illustration, imagine IRB deliberations regarding whether to ask participants if they want to learn about any heightened cancer susceptibility that their sequence data could reveal. There are two options, both of which involve accepting that a mistake will be made. If an IRB requires investigators to actively solicit preferences, it is almost certain that some people will choose not to receive their cancer predisposition information without fully understanding the implications of this choice. Perhaps subjects will make this choice because they do not understand the nature of the information they are refusing and/or because they fear the anxiety associated with confronting bad news. Whatever the reason, soliciting preferences means accepting that some people will not gain access to potentially life-saving treatment that they would have actually wanted had they been fully informed.²⁹ In contrast, if investigators do not actively solicit preferences, some people could be forced to confront the fact that they might battle cancer in the future, even though they would have legitimately preferred not to know this fact.³⁰

Rightly sensing a need for professional ethical guidance, the American College of Medical Genetics and Genomics (ACMG) published recommendations suggesting that there is an obligation to look for secondary findings, and that those findings be routinely returned without soliciting a patient’s preference for knowing or not knowing that information.³¹ These recommendations sparked an extended heated debate about the importance of the RNTK,³² eventually prompting the ACMG to revise their guidance to specify that patients must be given an opportunity to opt-out of receiving genetic information.³³ This RNTK debate was particularly heated in the research context, where open questions about the scope of an investigator’s clinical responsibilities towards their research subjects persist.

B. Newborn Blood Spot Research

A day or so after birth, a nurse will approach new parents about obtaining a tiny blood sample (typically from a heel stick) that can screen their newborn infant for a range of conditions that generally require early intervention.³⁴ While newborn screening is an important public health activity, the samples can also be extremely valuable from a medical research standpoint.³⁵ Since these newborn bloodspots (NBS) are taken from almost all newborns, they provide an opportunity to generate population-level data about public health questions and can be vital for ascertaining rare disease cases in sufficient numbers to conduct rigorous research.³⁶ NBSs are also uniquely useful in epidemiological research, particularly relating to infectious diseases and environmental exposures.³⁷ Traditionally, because these bloodspots are deidentified, research activities were not considered to be human subjects research under the Common Rule.³⁸ Thus, researchers have generally been allowed access to the samples without explicit informed consent. In the hours before being discharged from the hospital, new parents are generally given a brochure (along with a stack of other forms and pamphlets) that explains how the NBS research program works and provides information about the mechanism parents can use to opt-out.³⁹ One can reasonably question whether new parents have the cognitive capacity to fully engage with this kind of abstract question, buried in an avalanche of other more urgently important information, in the disorienting few days after delivery.

It is not surprising then that scholars and advocates have raised multiple concerns about the research use of NBS samples even though they hold significant scientific value. The primary argument is that it is unethical (and deceptive) to store samples for future research use without first having obtained explicit consent to do so.⁴⁰ There is concern about tangible harms flowing from the use of these samples⁴¹ and concerns about non-welfare harms (such as using samples to conduct controversial research that the source of the sample would not want to support consent to).⁴² These concerns have led to a number of high-profile court cases, where parents have successfully challenged NBS research programs that do not actively solicit consent.⁴³

The desire to obtain consent is understandable, but researchers have voiced concerns that imposing such a requirement would reduce participation rates, thus undermining the value of a resource as a broad representation of the population.⁴⁴ This has led professional societies and policymakers to issue a range of guidance positions. The American Society of Human Genetics (ASHG), in a 2015 update of their guidance on pediatric genetic testing, advocated for parents to be given a choice about future research use of NBS samples.⁴⁵ Even more radically, the ACMG and the American Academy of Pediatrics (AAP) jointly argued for the necessity of consent before any newborn screening (let alone research) is conducted. The AAP did, however, remain flexible about the form that consent process can take.⁴⁶ Other groups remain equivocal,⁴⁷ argue in favor of the status quo,⁴⁸ or endorse the acceptability of an opt-out approach.⁴⁹

C. Genomic Data Sharing

In 2015, the National Institutes of Health (NIH) implemented a genomic data sharing policy that applied to all NIH-funded research.⁵⁰ Any project that would generate large-scale genomic data was expected to deposit their data in a public repository that could be accessed by other researchers.⁵¹ The purpose of the NIH genomic data sharing (GDS) policy was to increase the amount of available genomic data, which in turn allows for more powerful analyses of the relationships between genetics and human health.⁵² NIH has a long history of encouraging the sharing of data, but the GDS represented an ambitious attempt to move towards the promise of personalized medicine.⁵³ Since most genetic traits are not mendelian, aggregation of large genomic data sets allows for researchers to explore the complex relationships between an array of interrelated genetic variants that can each contribute in subtle ways to a given health trait or disease.⁵⁴

With the advent of this policy, there has been extensive literature on balancing the maximization of data utility with the protection of human subjects from the risks of broad genomic data sharing.⁵⁵ If explicit consent is required, some people will decline to share, thus decreasing the breadth and value of the data.⁵⁶ Though there have been calls for tiered consent⁵⁷ (where subjects are given choices about how their data can be shared), there is also a recognition that it is sometimes acceptable to maximize the deposition of genomic data by not presenting an explicit choice about sharing one’s data.⁵⁸

Unlike the NBS consent debate, policy guidance related to implementation of the GDS policy called for explicit consent and rejected opt-out systems.⁵⁹ Even if a decision was made that CMNs were not appropriate when obtaining consent for broad sharing of genomic data, the debate about the contours of what is or is not ethically appropriate are relevant to our discussion. Policy makers attempted to navigate between the competing requirements of what law, ethics and public opinion required, while not unnecessarily limiting the scope of data sharing.⁶⁰

Though the GDS policy outlines an expectation that explicit consent will be obtained, this tension is illustrated by the NIH Office of Science Policy’s (OSP) suggestion that a number of consent features that leave open the possibility that explicit consent for the broad sharing of genomic data will not be necessary in some situations. For example, though explicit consent would be necessary for open access, OSP guidance argues that explicit consent is not necessary for data to be deposited in the controlled access tier.⁶¹ Similarly, when the original research project has been granted a waiver of informed consent or when the Common Rule regulations do not apply, mere disclosure of the data sharing plan may be sufficient and exceptions to the consent requirement can be sought.⁶²

In practice, it appears that research institutions have favored broad consent strategies.⁶³ One can imagine a range of possible consent approaches, from very granular to very passive.⁶⁴ On the granular end, research subjects could be presented with a detailed series of choices about how their data can be used (such as a tiered consent).⁶⁵ A less granular, but still explicit approach, includes providing details about the data sharing plan and giving an explicit choice about whether to opt-in.⁶⁶ A more passive approach includes asking for broad consent by including details about the data sharing plan in the consent process without drawing special attention to it or asking for consent about that discrete topic; general consent to participate would include agreement to broad data sharing.⁶⁷ According to one review of consent materials, granular, tiered consent appears uncommon, with the majority of research institutions obtaining one-time broad consent as part of the standard consent process (which does not include obtaining separate, explicit consent for data sharing).⁶⁸ The concept of asking for high-level rather than granular consent is therefore in line with the spirit of CMNs.

Though none of these debates were framed in terms of nudging, they all raise questions about the propriety of instituting an institution-level choice masking nudge. Is it ethically acceptable to nudge research participants towards receiving genetic information by hiding the existence of an opportunity to opt out of learning such information? Similarly, is it ethically acceptable to hide a choice to opt out of having your infant’s sample be stored for later research use, creating a default that maximizes the number of available research samples? And is it acceptable to require only broad consent in research, instead of requiring that participants agree to more granular options? Questions of this variety abound in medical research, and as the GDS case shows,⁶⁹ the text of regulations alone does not always clarify what should be done in practice. For that reason, we provide commentary on the applicability of the reasonable person standard, the legal standard applied to informed consent regulations to supplement our reading of the regulations.

Part III: Libertarian Paternalism and Choice Masking Nudges

To provide clarification on what a CMN is, as well as the potential ethical concerns such interventions might raise, it is useful to provide a brief introduction to nudging. Nudges were first proposed by Cass Sunstein and Richard Thaler as noncoercive interventions to improve individual wellbeing and decision-making, supported by their proposed doctrine of libertarian paternalism.⁷⁰ Drawing on insights from behavioral economics suggesting that human decision-making is plagued by systematic cognitive biases and irrational decision-making heuristics, nudges trigger or tap into these biases or heuristics to help people make better decisions for themselves.⁷¹ A nudge is “any aspect of the choice architecture that alters people’s behavior in a predictable way without forbidding any options or significantly changing their economic incentives.”⁷² This is distinct from more familiar methods of persuasion, incentives, and coercion. Proponents of nudging claim that it is particularly useful when decisions “are difficult, complex, and infrequent, and when they have poor feedback and few opportunities for learning,” and when choice architects can “make good guesses about what is best for the Nudgees”—two conditions likely to be met in medical research.⁷³

A classic example of a libertarian paternalist nudge is changing the location of food in a cafeteria so that healthy options are at eye-level and unhealthy ones are less conspicuous.⁷⁴ The idea here is that while many people claim they want to eat healthier, when the time comes to do so, they are more likely to opt for what is easily accessible and in front of them rather than choose what will promote these ends, especially when they are hungry and faced with several ready-made but unhealthy options.⁷⁵ In other words, although people value and care about healthy eating, they tend not to choose foods in ways that promote or align with these values, due, perhaps, to fast and unconscious decision heuristics or weakness of will.⁷⁶ Nudging is proposed to help people overcome those failures and make choices that better promote or align with their values by intentionally modifying the location of healthy foods in cafeterias. Since we know that people are likely to choose what is salient and easily accessible, we can exploit that simply by changing the items that are most prominent from, say, pastries and candy bars to apples and pears.⁷⁷ This nudge is libertarian in the sense that it is noncoercive. People are free to resist the nudge if they want to badly enough: unhealthy foods are still available to purchase at the cafeteria; they are just less salient.⁷⁸ But this nudge is also paternalistic in the sense that modifying the location of foods has the predictable and intended effect of directing people to make healthier choices.

Sunstein and Thaler claim that “to count as a mere nudge, the intervention must be easy and cheap to avoid.”⁷⁹ In order words, nudges must be easily resistible. Sunstein and Thaler distinguish nudges from other ways of influencing choice, such as providing information or attempting to persuade, as well as from more familiar instances of paternalism that coerce people to act in ways that promote their own interests by eliminating options or imposing significant costs on certain choices.⁸⁰ For example, laws that require the use of seatbelts in cars and helmets by those riding motorcycles are paternalistic; they are intended to promote or protect the interests of people who ride in cars or on motorcycles.⁸¹ But people are not free to disobey these laws without incurring significant penalties.⁸² Nudges involve a weaker form of paternalism, in that they are noncoercive and influence choice—not by actually removing certain options or attaching significant penalties to them, but rather, by altering the presentation of options in a way that influences choice in a predictable way. Recall our cafeteria example. The libertarian paternalist does not advocate for tripling the price of unhealthy items or ceasing to sell them at all. All she advocates for is changing the location of items to make certain options more or less salient to her hungry customers.

Some might claim that CMNs cause trouble for the “libertarian” part of libertarian paternalism. Aren’t nudges that mask available options significantly different than nudges that simply change the location or salience of options, since the former involve something like intentionally hiding available options while the latter does not? We aver that CMNs still qualify as “libertarian” (in the sense that they respect individuals’ autonomy and freedom to choose). While certain choices are masked and so not listed on the menu of options, masked options are still easily available to those who seek to access them, and the possibility of opting-out should be sufficiently clear to those who wish to exercise it. What is and is not being masked is crucial in our analysis of CMNs. To be defensible, information about masked options and how to access them must be easily available. Only then does a CMN meet the resistibility requirement.

Rearranging foods in a cafeteria so that people make healthier choices for themselves will strike most as an innocuous intervention. And indeed, the classic examples that proponents of nudges appeal to when introducing and motivating this novel policy tool tend to be similarly innocuous. Sunstein and Thaler talk about nudging more employees into saving for retirement by changing the default enrollment from opt-in to opt-out, or restaurants serving food on smaller plates to make diners believe they are eating more than they actually are.⁸³ But some proposed uses of nudges are more controversial. Such controversial nudges extend beyond low-stakes decision environments like cafeterias and into high-stakes decision environments like exam rooms, where patients might be nudged into choosing a treatment by a physician who frames the risks and benefits of that treatment in a particular way.⁸⁴ Moreover, not all proposed uses of nudges are intended to benefit the individual who is nudged. For example, consider opt-out systems for organ donation that default drivers into becoming an organ donor—increasing the supply of organs surely benefits society as a whole, but it is hard to see how such a nudge could benefit the donor directly. Exactly who benefits from nudges is an important issue and in what follows, we will argue that nudges that are intended to benefit the individual who is nudged are easier to justify than nudges that are supposed to benefit someone else, or society as a whole. All three circumstances, however, are justifiable, provided they meet our acceptability conditions.

The ubiquity of nudging makes sense once we recognize that the features of decision making that make us susceptible to nudges are present whenever we make decisions. We cannot help but be influenced by the way things are presented to us, nor can we help but interact with a world that is presented in one way rather than another. Sunstein and Thaler term this the “inevitability” of choice architecture and its influence: people have to make decisions, and it is not as if there is some neutral way of arranging things that will have no effect on the decisions people ultimately make.⁸⁵ There is no such thing as a neutral way of presenting or framing choices, since how choices are presented or framed will inevitably affect what people choose. If those who are in a position to arrange choices cannot help but to exert some influence, then does it not make most sense to arrange things in a way that is likely to steer people towards good choices and away from bad ones? Indeed, this appeal to the inevitability of influencing is often a powerful response to those who argue that nudges are objectionably manipulative.⁸⁶

Even if we cannot help but influence peoples’ decisions in some way, and even if nudges are noncoercive and so are in this sense less objectionable than strongly paternalistic interventions, nudges still involve an intentional manipulation of the environment: choice architects steer people toward certain choices and away from others. This sort of meddling cries out to many as standing in need of justification. Thankfully, proponents of nudges have had a lot to say in their defense.

First, as we have seen, nudges promote beneficial outcomes. Effectively implemented nudges help people make choices that make benefit them: they eat healthier, save for their retirement, enroll in health insurance, and make contributions to their pension plan.⁸⁷ Many people prefer these things for themselves but will not take action to bring them about due to the operation of cognitive biases or heuristics. Effectively implemented nudges also help people make choices that make other people better off: they become organ donors and safer drivers, they donate to charity, they prescribe their patients generic versions of drugs rather than the more expensive brand names and speak to their patients about their goals and values for end-of-life care.⁸⁸ And finally, effectively implemented nudges also help people make choices that promote public goods: they litter less, recycle more, and use less energy.⁸⁹ Nudges can play an important role in helping to bring about these important beneficial outcomes.

Second, nudges respect freedom of choice. Nudges influence choice, not by changing the menu of available options or by significantly changing economic incentives by making some available options significantly more expensive than others, but rather by changing the way options are presented or framed, thus respecting freedom of choice.⁹⁰

Finally, survey evidence suggests that people do not mind nudges and actually prefer nudges to other, stronger forms of influencing choice, at least when they endorse the goals the nudges are intended to promote.⁹¹ For example, there is survey evidence illustrating that in liberal democracies, strong majorities approve of nudges that protect and promote health and safety, as well as those that protect the environment, and that they prefer these to more coercive laws that promote those same goals.⁹² Interestingly, the hypothetical nudges that individuals were surveyed about generally concerned public health outcomes—such as efforts to prevent obesity, limit salt consumption, discourage tobacco use, and reduce deaths from distracted driving.⁹³

Part IV: The Reasonable Person Standard and Research Regulations

The proposal to use CMNs in medical research might initially appear to directly contradict the requirements of full disclosure that came to dominate the informed consent jurisprudence in clinical care and medical research in the mid-twentieth century. However, we have reason to believe that CMNs are compatible with existing jurisprudence and legal norms. Our arguments depend on two claims: first, that there is a key difference between choices that have significant bodily autonomy implications and those that are primarily informational; and second, that disclosure requirements do not necessarily require the presentation of explicit choices about the information disclosed. In this section, we argue that the aim of the reasonable person standard, properly understood and applied to the kind of informational decisions in medical research that we are concerned with here, supports CMNs (even if it may have legitimately been taken to support full disclosure requirements in other contexts). Because this standard is part of the United States federal regulations governing research and because the reasonable person standard is so deeply embedded in the American legal tradition,⁹⁴ a justification for CMNs that appeals to the reasonable person standard also fits into that tradition.⁹⁵

A. Informed Consent Jurisprudence Supporting Full Disclosure

Informed consent jurisprudence from the second half of the twentieth century is characterized by landmark decisions underscoring the importance of physicians disclosing relevant medical information to their patients.⁹⁶ The cases that make up that jurisprudence typically involved physicians failing to inform patients of bodily harm risks that subsequently materialized.⁹⁷ In early informed consent cases, the patients’ right to full disclosure was ostensibly grounded in their right to bodily integrity.⁹⁸ That autonomy interest was sufficiently important to ground the right to be informed about actions that may infringe it, and to be presented with an option to not participate in invasive medical procedures. But the right to bodily integrity can be meaningfully analyzed in terms of the reasonable person standard (RPS), as later informed consent rulings have done.⁹⁹ Intuitively, a reasonable person would expect full disclosure of relevant treatment options and risks in a situation where her bodily autonomy is at stake.¹⁰⁰ A standard of reasonableness would also give a research participant claims to a high level of explicit choice regarding decisions with bodily autonomy implications. Understanding the informed consent requirements in terms of the RPS is helpful because it offers guidance even in cases of research questions that do not obviously pose threats to bodily autonomy, as is the case for many of the informational choices we consider in this paper as candidates for CMNs.

B. The Reasonable Person Standard and Medical Research Nudging

Before returning to our argument regarding the RPS and CMNs, it is worth making some more remarks about the reasonable person standard. The difficulty of creating rules that will be appropriate in every context has been long acknowledged.¹⁰¹ The RPS responds to this challenge by building flexibility into rules so their guidance may remain relevant, despite changing social contexts. In fact, the content of the RPS cannot be precisely pinned down by law because it refers to social facts beyond the reach of legislation.¹⁰² The standard is widely accepted in a variety of legal applications, including negligence, contract, administrative law, judicial review, and criminal law.¹⁰³ A typical definition of the reasonable person standard describes “an imaginary actor who represents the community consensus of acceptable or appropriate behavior. This consensus establishes neither a standard of what average persons do nor an aspirational ideal beyond the reach of most persons, but a minimum threshold below which the ordinary person may not fall without being found deficient.”¹⁰⁴ Despite some disagreement on the nature of the RPS,¹⁰⁵ experts generally agree that the standard must be applied on a case-by-case basis¹⁰⁶ and that it includes a range of behavioral standards.¹⁰⁷ Ultimately, the RPS provides instructions on making judgements about what a reasonable person would do, and its legal prominence gives those judgements legal weight. Standards set by the RPS, formulated in this impersonal way, “do not bend to the varying personal characteristics of those who are judged by them.”¹⁰⁸

Nevertheless, describing how the RPS is used does not tell us what it means to be reasonable. One suggestion from philosophical literature comes from Thomas Scanlon, who notes that a reasonable person must both consider an appropriate body of facts to be relevant to the question at hand and engage in an acceptable pattern of reasoning on the basis of those facts.¹⁰⁹ This process should occur against the backdrop of acceptable general aims and concerns.¹¹⁰ Importantly, reasonable does not mean self-interested; a reasonable person may sometimes recognize that others’ interests override her own.¹¹¹ The reasonable person must have reasonable concerns and respond to them in a rational way. For our purposes in analyzing CMNs, this standard imposes some normative constraints on reasonable research participants. While they can and should be concerned with their own welfare and utility, research participants should also be responsive to social interests and to the interests of others, and the way they make decisions should reflect those considerations. Their reasoning should give appropriate—but not overriding—weight to social interest, and they should also strive to act in ways that they can justify. Fleshed out in this way, we have a somewhat clearer picture of what it means to apply a reasonable person standard to participants in medical research. And it should be unsurprising that as context—that is, the reasons one must respond to—changes, the most reasonable courses of action may change as well.

The RPS is helpful in explaining why the aforementioned body of jurisprudence that stresses the importance of informed consent may not translate directly to some contemporary medical research situations. The key contextual difference comes from the type of decision participants are asked to make in the research setting—specifically, whether they have immediate implications for bodily integrity or whether they are primarily informational.¹¹² Informational choices are less likely to infringe on a research participant’s bodily integrity and more likely to affect individuals other than the nudgee than are physical choices, which primarily affect the research participant in an immediate bodily way. Potential benefit to others provides further support for informational nudges that are unlikely to be available for physical nudges. Reasonable disclosure and consent expectations where the physical consequences of medical research decisions are severe and individual decisions are not likely to significantly affect others, we argue, differ from reasonable expectations in contexts where decisions carry minor risk of causing serious injury but could have important consequences for research participants, others, and society more broadly. The RPS would not necessarily require explicit disclosure of the choice hidden by CMNs for two reasons: first, the CMN should be accompanied by sufficient information for participants to realize they have the option of opting out of defaults and for them to make sense of the defaults they are being nudged toward; and second, the fact that a choice is masked is expected to be in the participant’s best interests.

C. Regulations and the RPS

If the RPS can theoretically accommodate CMNs, the next question is whether existing regulations governing research with human subjects legally permit their use. In the United States, the revised Common Rule lays out detailed requirements for informed consent.¹¹³ These regulations are interesting for our purposes because in addition to referring to RPS, they also seem to allow for CMNs, provided that the CMNs do not also mask important information.¹¹⁴ The most general level of guidance from the Common Rule focuses on ensuring that research participants are provided with enough information and opportunities for discussion to make informed decisions.¹¹⁵ Note that the regulations do not specify how that information be presented, or what kinds of choices must be made available, much less clearly communicated:¹¹⁶

1. (4) The prospective subject or the legally authorized representative must be provided with the information that a reasonable person would want to have in order to make an informed decision about whether to participate, and an opportunity to discuss that information….

   1. Informed consent must begin with a concise and focused presentation of the key information that is most likely to assist a prospective subject or legally authorized representative in understanding the reasons why one might or might not want to participate in the research. This part of the informed consent must be organized and presented in a way that facilitates comprehension.
   2. Informed consent as a whole must present information in sufficient detail relating to the research, and must be organized and presented in a way that does not merely provide lists of isolated facts but rather facilitates the prospective subject’s or legally authorized representative’s understanding of the reasons why one might or might not want to participate.¹¹⁷

The general guidance regarding information and understanding is followed by a more specific list of the kinds of statements and disclosures to be provided to research subjects.¹¹⁸ Again, the regulations focus specifically on providing information in the form of statements and do not mention soliciting choices.¹¹⁹ The regulations are oriented toward disclosing enough information for research participants to adequately understand what they are agreeing to participate in. One of our conditions on CMNs is that they must provide sufficient information to understand the implications of the default decision.¹²⁰ While CMNs involve masking some choices—specifically, by only presenting participants with a set of default options outlined in the consent form—they may not hide information relevant to the choice. In fact, providing such information is necessary to ensure that the default choice opt-out mechanism is robust. Since opt-out mechanisms rely on participants’ self-identifying concerns with default options, they must be provided with enough information to be able to form an opinion about those options and be made aware that they can voice disagreement or discomfort at any time.¹²¹ Since the informed consent regulatory requirements are silent on whether additional explicit choices (i.e., other than the primary decision to enroll in a study) must be presented as long as there is reasonable disclosure, it seems entirely compatible with our presentation of CMNs. Ultimately, the RPS must be used to determine what qualifies as sufficient information. The revised Common Rule explicitly refers to the RPS because it is the standard-setter used to put these regulations into practice.¹²²

D. Making Use of the RPS

A plausible picture of how the RPS could be applied to CMNs, consistent with prevailing United States research regulations, emerges from the above discussion. If we accept that a reasonable choice is one that balances interests and harms in a reasonable manner, we might accept something like the following: medical researchers must present research participants with the choices that a reasonable person would have an interest in being able to make. That stipulation prevents researchers from masking certain choices—specifically, ones ruled out by the autonomy and acceptability considerations articulated later in this paper—although it can at best be used as a rough heuristic, given the controversy over the definition of reasonableness. By the same token, we would also accept the similar claim that medical researchers are not under an obligation to provide research participants with choices that no reasonable person could claim a right to make. Such a position makes it clear that researchers have some discretion in which choices they present to research participants. Notably, while we understand the RPS as justifying masking the number of options presented to research participants, we do not take it to justify information masking. The balance of individual, collective, and social consequences of masking choices explain why it is reasonable to do so. In the situations we are interested in, masking information would not contribute to those positive consequences (and would infringe upon important autonomy interests), and thus would not be supported by the RPS.

Part IV: Autonomy and Self-Determination

While bioethicists all agree that autonomy is an important ethical consideration, there are many different views about what autonomy is and what type of conduct violates it.¹²³ For our purposes, a rough and fairly intuitive characterization will suffice: autonomy is the capacity to be one’s own person or to be the author of one’s own life.¹²⁴ A person exercises her autonomy when her choices reflect, are based on, or align with, reasons, values, or preferences that are her own, rather than those that have been externally imposed on her.¹²⁵ We respect someone’s autonomy when we let them choose for themselves, according to the things that matter to them, and accept the decisions that they make. The aim of this section is not to challenge the preeminence of autonomy or to endorse any particular account of it, but rather to argue that CMNs are at least compatible with many plausible views of autonomy and may even promote it. Ultimately, we maintain that whether a CMN is justified depends on a comparison of the benefits to be gained by nudging with the autonomy-costs that come from masking the choice. Relevant to this question, we shall argue, is whether the CMN satisfies the four conditions we lay out in Part VI. In our view, when these conditions are met, the benefits to be gained by masking a choice are sufficient to justify the cost to autonomy. Our primary concern is with helping research participants make good decisions that they care about, which can itself enhance autonomy.¹²⁶

In this section, we will survey prominent accounts of what autonomy is and why we must respect it and argue that most of these views are compatible with CMNs. We acknowledge, however, that some accounts of autonomy sit less comfortably with CMNs. Specifically, CMNs might cause trouble for views on which respecting autonomy requires engaging with a chooser’s deliberative capacities or ensuring that a chooser decides on the basis of information that is relevant to the choice. People who believe that respect for autonomy requires interacting with a chooser’s decision-making capacities in this particular way are unlikely to accept our defense of CMNs. However, we suggest that the relevant question is not whether there is any cost to autonomy at all, but rather how significant that cost is, and whether the stakes of the choice are great enough that such an autonomy-violation is ethically unacceptable.

Though we defend CMNs in a variety of research settings and for a variety of decisions, we recognize the existence of a special domain of decisions that relate directly or are foundational to one’s identity or deep moral commitments, and we claim that individuals ought not to be nudged regarding those decisions.¹²⁷ Consider, for example, a sincerely held religious belief about what makes life worth living or the sort of activities one ought never to engage in. Acting in accordance with those kinds of deep commitments fosters autonomous authenticity.¹²⁸ CMNs that nudge individuals to make decisions that go against those commitments, if effective, could make them feel complicit in something they consider morally reprehensible, resulting in feelings of moral distress and autonomy harm.¹²⁹ For example, imagine that someone who, for religious reasons, is morally opposed to genetic cloning is nudged into providing genetic material to a biobank primarily used to support research into cloning. Because her deep religious commitment against cloning is so central to her identity, the sense of moral harm from being nudged into contradicting it could constitute an autonomy harm. Plausible views of autonomy recognize this protected domain; the domain thus constrains the kinds of choices that we may permissibly mask.¹³⁰

Even if the idea that such a protected domain exists makes sense in theory, however, respecting that domain poses serious practical challenges. For one thing, how can we evaluate the sincerity of individuals’ identity-centric convictions? One individual’s deep moral commitment is another’s weak preference. Relatedly, how are we to respond to extremely uncommon religious beliefs or moral commitments that researchers are unlikely to anticipate? Some heuristics may reduce the extent of this problem. For example, it is easier to anticipate the deep convictions of groups than of individuals, because groups often publicly articulate their fundamental commitments. Self-identified group membership is thus a likely indicator of beliefs that fall within the protected domain, albeit an imperfect one. For some decisions, public survey data may be helpful in judging whether a particular view is common and deeply held—some surveys, for example, show that many people express deep discomfort with research into genetic cloning.¹³¹ Researchers should make their best efforts to anticipate what kinds of CMNs are likely to involve nudging people into identity-formational decisions and consequently, should strive to avoid utilizing those CMNs. The unfortunate fact that researchers may not always succeed doing so underscores the importance of ensuring that CMNs only mask information—not choices—and that easy opt-out mechanisms exist.

A. Genuine Choice

A prominent feature of many views of autonomy is that they require an autonomous individual to possess a genuine choice between a variety of different outcomes.¹³² While few views on autonomy hold that genuine choice is entirely constitutive of autonomy, it is so often a prerequisite for autonomy that it warrants discussion on its own.¹³³ Genuine choice has two distinct components: availability of options and the freedom to choose between them.¹³⁴ A critic of CMNs might claim that these nudges should be rejected since they restrict both the set of available options and someone’s freedom to choose from that set. If true, this objection would constitute a serious strike against the argument for using CMNs.

As is clear from their very name, CMNs attempt to make certain choices less apparent to those being nudged. Recall that the primary motivation for introducing CMNs was to not suggest certain options to individuals who did not have prior preferences about those options when those options are expected to have negative effects on the chooser or on others.¹³⁵ CMNs are not intended to actually reduce the chooser’s option set—importantly, they must be easily resistible and include a clear opt-out mechanism—but they are undeniably intended to make the option set appear more limited.¹³⁶ The opt-out mechanism must be easy to locate for individuals who wish to choose a masked option. But what about the individuals who never seek an opt-out mechanism, regardless of whether they would have wanted it, because they do not realize that an option was masked in the first place? For individuals who recognize the CMN and take the easy, available steps to opt-out, the nudge clearly does not interfere with either component of genuine choice. The important question, therefore, is how significantly masking a) decreases the number of available options and b) interferes with the ability—for individuals who do not act to avoid the nudge—to choose.

It seems unlikely that CMNs will affect the number of available options or research participants’ ability to choose to such a degree as to make their choice nonautonomous. This claim may ultimately result in the need to compare the costs and benefits of the CMN and its effects on genuine choice. As we emphasized above, the sort of CMNs we aim to defend affect only the appearance of choice: to be justified, a CMN must not alter the underlying options and must be easily avoidable. In practice, the nudgee must be aware that some choices are being masked, and an easy way to avoid the nudge must be available if they seek it. Thus, we contend that if ethically acceptable CMNs restrict autonomy, this is only a minor restriction. And importantly, ensuring a maximal degree of autonomous choice is not the only desiderata for a CMN; it is also important that the nudge helps participants make better decisions in medical research. Thus, even if CMNs unavoidably represent some deviation from the genuine choice conception of autonomy, the safeguards that we stipulate ensure that they protect autonomy to a sufficient degree. When we compare the potential for a (slight) diminution of genuine choice with the benefits expected from an acceptable nudge, the CMNs seem justified.

B. Making Important Decisions

CMNs might also be criticized on the autonomy grounds that they prevent research participants from making important decisions for themselves, since being able to make certain important decisions is a prominent feature of some accounts of the ethical importance of autonomy.¹³⁷ On these views, what matters for autonomy is that the person makes the important choices that clearly shape the major features of their life. But not all choices are equally important in shaping a life. As we explained earlier in this section, we should not mask choices that relate to the protected domain of identity-related decisions; those kinds of choices are undoubtedly life-shaping.¹³⁸

Making important decisions does not require that autonomous people have genuine choice over every decision they could possibly make. While respect for autonomy requires allowing individuals to make the most personal identity-forming decisions independently, many decisions in medical research are not of that type.¹³⁹ Just as we routinely accept that autonomy interests do not allow us to make all types of decisions in our everyday lives— such as which side of the street to drive on, whether to educate our children, and what we pay in taxes—autonomy does not require that we make all types of decisions in medical research. Some constraints on choice are a necessary part of living in society and do not unduly impinge on autonomy.¹⁴⁰ Interestingly, this account of what matters for autonomy might supplement arguments for the use of CMNs.¹⁴¹ Nudges that mask relatively minor choices could actually promote autonomy, then, by freeing individuals to expend their limited cognitive energy on decisions that are the most crucial to personal autonomy.

C. Understanding and Engaging with Options

Another way to think about how CMNs might infringe upon autonomy comes from the requirements of informed consent, a widely recognized and core tenet of medical ethics intended to ensure that research participants understand what they are agreeing to. Roughly, the idea is that obtaining informed consent is a good way to ensure that a participant’s decision is autonomous— but only if the participant has sufficient understanding of the decision. Adequate understanding is one of the key requirements for informed consent outlined in Beauchamp and Childress’s famous text outlining the basic principles of bioethics.¹⁴² On their view, autonomous choice requires not only a sufficient range of options, but also sufficient understanding of those options, as well the cognitive capacities that are necessary to make well-informed decisions on the basis of that understanding.¹⁴³ Further, one might add that it is not enough merely to understand the options and have the necessary cognitive capacities to make well-informed decisions; one must also engage those cognitive capacities, ensuring that the chooser deliberates about the options prior to making a selection.

This view of autonomy is concerned with the process of making a decision, and CMNs are admittedly more concerned with decision outcomes than with processes.¹⁴⁴ CMNs are not designed to be educative; their aim is not to inform or increase one’s understanding, nor is it to make it more likely that people choose on the basis of relevant information.¹⁴⁵ Rather, CNMs aim to help people make better choices, specifically by steering them away from choices that researchers have good reason to believe are bad.¹⁴⁶ So, if autonomy requires an understanding of the options that are available, being well-informed about the information that is relevant to a choice, deliberating, and ultimately choosing an option based on that relevant information, CMNs may threaten to infringe on autonomy because they cannot ensure that participants understand the default decisions they are nudged toward.

However, it is important to notice that this is not a knock-down argument against CMNs. For one thing, it is not obvious that slightly reducing people’s information counts as a real cost to their autonomy. We typically do not believe that people always need to be informed about all the available options in order to act autonomously: what matters is that they are sufficiently informed. So, for example, we do not think that a clinical investigator infringes on the autonomy of their subjects by failing to tell their subjects the entire range of information that might somehow bear on their choice about whether to participate. Indeed, it is hard to see how a clinical investigator could do that and still have time to carry out any of their research. Rather, the investigator has an obligation to disclose all the information that they have reason to believe would be relevant to the prospective participant’s decision.¹⁴⁷ The fact that a specific CMN exists may not be necessary to disclose if participants are informed that they may request modified treatment regarding default options they disagree with.¹⁴⁸ But, even if one resists this argument and thinks that providing less information always has some autonomy cost, the cost is likely to be quite low in the sorts of cases we are concerned with.¹⁴⁹ We are not proposing an insuperable barrier to people receiving the relevant information about alternative options, but we are requiring that they take an extra step. That is why CMNs are nudges, and not more stringent restrictions on choice. Since this cost is quite small, it is likely to be outweighed in most of the cases we are concerned with, either by the benefits of the nudge, or even other considerations of autonomy—for example, the fact that CMNs, when well-executed, can lead people to make choices that better align with their own values.¹⁵⁰

D. Successfully Reaching One’s Own Goals

A final way to understand autonomy draws on the distinction made between “means” and “ends” paternalism. Ends paternalism involves guiding people toward goals that they do not identify as their own, whereas means paternalism involves guiding people toward goals that they do identify as their own.¹⁵¹ Put slightly different, means paternalism assists people in taking the correct means to their own ends, whereas ends paternalism involves changing people’s ends.¹⁵² In his book, Why Nudge?, Cass Sunstein argues that we respect autonomy so long as we respect people’s own ends.¹⁵³ In other words, as long as nudges are always instances of means paternalism, and never instances of ends paternalism, then we need not worry about whether those nudges respect autonomy because respecting autonomy is merely about “allowing individuals to make informed decisions about their own ends.”¹⁵⁴ According to this view, in order to assess a nudge’s impact on autonomy, all we must ask is whether the person who is nudged is “better off, as judged by themselves.”¹⁵⁵ We decide autonomously when we ultimately get what we want, and we respect someone’s autonomy when we help them get what they want.¹⁵⁶

There are reasons to doubt that this account of autonomy really captures what it is to be autonomous, and therefore what it is to respect someone’s autonomy. For one thing, the distinction between means and ends paternalism is not as sharp as this view might suggest. In many cases, participants will not have strong prior preferences about the options being masked or towards which they are being nudged.¹⁵⁷ Moreover, individuals adopt ends for a variety of reasons, including because of previous nudges or other social influences.¹⁵⁸ Self-proclaimed ends may reflect prior influences just as much as they reflect what is central to an individual’s autonomy.¹⁵⁹ It is thus often unclear when nudges are more properly considered ends versus means paternalistic.¹⁶⁰ As a result, this view of autonomy is only of limited usefulness in analyzing nudges.

But leaving aside those concerns, it seems plausible that at least some CMNs are means paternalist and are therefore compatible with this view of autonomy. Returning to the RNTK case described in the introduction, consider a nudge that masks the choice of whether or not to receive actionable secondary genetic findings.¹⁶¹ While most people would prefer to avoid becoming incurably ill, when presented with a choice, some will elect not to receive these secondary findings, even though it may not align with their values and they may not choose to do so upon deeper reflection.¹⁶² Masking this choice, then, would lead more people to receive secondary findings and would benefit most people. As long as we can be confident that masking a choice will significantly decrease the chance that people choose badly—that is, choose an option that will make them worse off, as judged by themselves—then doing so might not disrespect autonomy. In fact, nudges could even turn out to be autonomy enhancing, on some views.¹⁶³ If autonomy is respected when people make choices that reflect their own values—and since people sometimes predictably choose in ways that do not promote their own values—then nudges could be considered autonomy-enhancing if and when they help a nudgee choose in accordance with their own values.¹⁶⁴

This approach raises a number of questions, however; most notably including how to respond to the minority of individuals on the tail end of the curve who genuinely do not identify with the aims of the nudge.¹⁶⁵ Is the fact that a nudge is means paternalistic for most individuals enough for a nudge to be considered means paternalistic overall? Even if there are some things that almost everyone values, people vary tremendously in the trade-offs they are willing to make. Taking this inevitable variation seriously should lead us to be skeptical about the idea of a “one-size-fits-all” nudge that somehow manages to be sensitive to this variation and steers people toward choices that facilitate them getting what they really want.¹⁶⁶ Exactly how is a choice architect able to access this information about each person whose choice is affected by the nudge, especially when they do not themselves have self-knowledge about what they really want or what would really make them better off? These are deep and interesting questions, and they raise issues for many of the views of autonomy discussed here. Our hope is that the acceptability conditions for CMNs that we outline in the next section help to answer them, at least for the nudges we aim to defend.¹⁶⁷

This section described commonly held beliefs about the importance of autonomy and argued that as long as CMNs do not intrude on protected identity-forming spheres, they are compatible with respecting individual autonomy. However, showing that compatibility is not yet enough to justify them. The next section takes up the question of what further features of CMNs are necessary to make them acceptable, particularly in the research context.

Part VI: Non-Autonomy Acceptability Considerations for Choice-Masking Nudges

While autonomy is an important consideration relevant to justifying the kind of informational nudges we are primarily concerned with here, it is not the only one.¹⁶⁸ In this section, we focus on four further considerations, all of which must also be met for a CMN to be ethically acceptable.¹⁶⁹ Acceptability comes on a continuum, and nudges toward certain types of research—such as research with vulnerable populations or research on ethically controversial topics such as cloning—may need to satisfy especially stringent acceptability conditions.¹⁷⁰ While the stringency of each consideration will vary with context, the following four considerations give a rough indication of how acceptable any CMN is likely to be. First, the nudge must be in the service of legitimate policy goals. Second, the nudge must result in significantly more benefits than harms, overall. Third, while nudges may have associated burdens, those burdens cannot be distributed unfairly. And fourth, for a CMN to be acceptable, it must be better than other feasible alternative ways of achieving the same goals. There are two components to CMN acceptability: first, the nudge must not be objectionable on autonomy grounds and second, it must satisfy the four considerations described below.

A. Legitimate Policy Goals:171

There are three levels on which a CMN policy goal can be legitimate: individual, collective, and societal. The idea that health promotion has a special status because of health’s fundamental role in furthering human activities colors the discussion of CMNs policy goals at all levels.¹⁷² That important role, we argue, is sufficient to ground a legitimate interest in promoting health—and, by extension, medical research.¹⁷³ First, on an individual level, it could be intended to benefit nudgees since they, themselves would endorse or otherwise approve of the underlying goal. Second, there is a broader collective basis of legitimacy goals for CMNs; some policy goals might be worth pursuing even if not every individual nudged by them would agree that the nudge makes them better off. CMNs can be justified on a collective level when most individuals who will be nudged agree with the goals that the nudge is intended to promote. Finally, CMNs can be justified on a societal level when they lead to consequences that benefit society or when they express important social values and attitudes. Though all three levels can provide sufficient justification for a CMN, as one gets further away from direct individual benefit, the threshold for justification increases.

Policy goals that are legitimate because of expected individual benefit are relatively straightforward to analyze. The most direct benefits from a CMN come from encouraging research participants to make decisions that they otherwise would not have made, and that have positive consequences.¹⁷⁴ For example, consider a participant who is nudged away from choosing to refuse secondary findings and who as a result, learns that her genome sequence contains evidence of high risk for Hereditary Non-Polyposis Colon Cancer, which she later develops.¹⁷⁵ After receiving information about her hereditary risk, she would be more likely to undergo enhanced routine screenings which would result in detection and treatment of the cancer at an early stage.¹⁷⁶ Without the genetic indication, the cancer, which is difficult to detect with only normal colonoscopies, would likely not be detected until a later and more fatal stage.¹⁷⁷ Thus, the nudge could yield a direct benefit to the nudgee herself.

In these cases, a CMN can steer someone towards a potentially life-saving decision that they otherwise might not have made but would have chosen if they had all necessary information and were perfectly rational. This is not a hypothetical concern: there is evidence that people regularly choose not to learn about secondary findings without fully understanding what they are actually refusing. Consider data from a recent empirical study on the RNTK in which a large environmental health study asked 8,400 participants if they would like to receive secondary genetic findings (“SFs”); 165 declined.¹⁷⁸ In a later sub-study, these “initial refusers” were given slightly more information about SFs and were given another opportunity to make a decision.¹⁷⁹ The results were stark, indicating a sizeable group of “weak refusers”:

…almost half of participants who initially refused SFs subsequently accepted them. By soliciting preferences through check boxes after an accurate but limited presentation of information, it is likely that some participants will make a choice that results in forgoing potentially life-saving information that, upon further reflection, they would have wanted to receive.¹⁸⁰

The decision to give participants a choice about whether to accept SFs introduced a risk: some participants might refuse to receive them. In such a scenario, a CMN that defaults participants into receiving these findings would mitigate that risk by guiding weak refusers towards a decision that is consistent with their own preferences and interests.¹⁸¹

On a collective level, a CMN can be justified if it benefits most participants. Benefits should be recognized and taken into consideration even when they accrue to someone other than the individual being nudged, or when the beneficiaries of the nudge cannot be directly identified. CMNs are carried out on an institutional level, not at the level of individual interactions.¹⁸² Thus, it will often not be feasible to nudge only some members of a targeted group; a nudge has the potential to affect the decisions of all who interact with it—those it will benefit and those it will not. It is enough to show a benefit if implementation of a CMN makes it more likely that most participants will act in accordance with the nudge; it is not necessary that the nudge is beneficial from the perspective of each individual being nudged. For instance, when we consider the RNTK case, we see that, although most participants would prefer to avoid becoming incurably ill, some participants might have principled reasons for deciding to opt-out of receiving SFs. For example, some people might worry that documentation of a potentially pathogenic genetic variant will make it harder for them to obtain certain kinds of insurance (such as health, life, and long-term care). If it is not possible to select only amenable nudgees, then applying the nudge to its intended audience will also involve applying the nudge to individuals outside of its intended audience.¹⁸³ We argue that nudging individuals who would prefer not to be nudged can be justified by the benefits that the nudge brings to others who are amenable to the nudge.

On a societal level, nudges may have legitimate policy goals even when nudgees do not have preferences about the goals (and would not, even if fully informed), provided that the goals serve an important societal purpose.¹⁸⁴ For example, current genetic reference data skews heavily toward white individuals of European ancestry.¹⁸⁵ As a result, genetic research findings may be less applicable to non-white individuals.¹⁸⁶ If genetic reference data were more broadly representative, it would have more value for more of the global population. The social good of having a diverse bank of genetic reference data could be a legitimate policy goal and justify a choice-masking nudge that urges individuals to contribute to such reference data banks, even if donors of that information are not likely to be directly benefited by the research it enables. Many social goals that will be valuable to large groups of people, now or in the future, do not directly benefit every individual who could help promote those goals. Important social goals are still legitimate to pursue and we see no principled reason why nudges cannot be used to promote them. When the individuals being nudged do not have strong preferences about the policy goal being promoted by the nudge, the societal legitimacy of the goal plays a greater role in justifying the policy goal.¹⁸⁷

Some people might worry that using legitimate policy goals as a justification for paternalistic CMNs is a dangerous move that opens the door to using CMNs in service of any frivolous policy aim that can be shown to be means-paternalistic¹⁸⁸ to most affected individuals or to promote a valid social goal that people do not oppose. This concern shows that policy aims of CMNs could become increasingly difficult to justify the farther away they move from providing direct individual benefits.¹⁸⁹ In response to these concerns, we emphasize that legitimate policy goals are a necessary, but not sufficient, condition for acceptable nudges. The considerations described in the remainder of this section deal with concerns about using nudges for trivial reasons.

B. The Benefits Outweigh the Harms:

In assessing the balance of benefits and harms from a CMN, we are interested in the direct and indirect effects on all people potentially impacted by the nudge. In the previous section, we began outlining the idea that the justificatory threshold will increase as the consequences of the nudge become further removed from the nudged individual.¹⁹⁰ For purposes of assessing benefits and harms, consequences that accrue directly to the research participant being nudged (the individual level) should be given the most weight. Broader social benefits and harms (the societal level) should be given less weight to reflect the less direct link between the nudgee’s decision and the ultimate goal, and because any individual’s marginal contribution to the larger social project will be less significant. Benefits and harms to third parties (the collective level) will fall somewhere in between, and the weight in these cases will depend on how direct the connection is between the CMN and the effect on the third-party. For a nudge to be acceptable, the aggregate benefits— weighted by their proximity to the individuals being nudged—must be greater than the similarly weighted aggregate harms.¹⁹¹

The various types of benefits that follow CMNs are described in the subsection above as the bases for different levels of legitimate policy goals.¹⁹² If a particular aspect of medical research can be reasonably expected to promote health (individually, collectively, or societally), there is at least a presumptive expectation that nudges that advance research have some important benefits.¹⁹³ We are concerned here with potential CMN-related harms of three types: economic, psychosocial, and harm to trust.¹⁹⁴ While research suggests that the importance of the former two concerns may be exaggerated, harm to trust in medical research is concerning—both in itself and because of its potential impact on health-seeking behavior and subsequent negative effects on health outcomes.¹⁹⁵ Complicating matters, the risk-benefit analysis becomes less certain when examining contexts like NBS and GDS, where the CMN produces less direct societal benefits.¹⁹⁶ The marginal contribution of any individual sample or data set will be minimal, but the aggregation of these resources can result in substantial benefit.

A final category of harm that we will not discuss here is harm associated with masking an important identity-foundational choice, whose forced selection could significantly undermine a nudged research participant’s sense of self. Such harms are unlikely to occur from choice-masking nudges that refrain from nudging around sensitive identity-foundational areas, as we stipulate that CMNs must.¹⁹⁷ It is possible, however, that even nudges that avoid masking most sensitive choices may have this kind of effect on a small number of participants. That potential harm can be minimized by making the CMN resistible and by flagging—as clearly as possible—the underlying conditions or beliefs that might make a particular nudged option inappropriate. It is unlikely that nudges satisfying all four acceptability considerations laid out in this section will lead to serious harms.

1. Economic and Psychosocial Harms

Economic and psychosocial harms are adverse consequences that should be avoided, but we are skeptical that otherwise acceptable CMNs would likely lead to serious harms of either kind. To illustrate potential economic and psychosocial harms that could come from CMNs, consider the RNTK empirical study mentioned above. While half of the initial refusers turned out to be “weak refusers”, the other half of the study’s cohort continued to utilize the opt-out mechanism.¹⁹⁸ Given that these “strong refusers” dissented twice and had a high level of understanding about the kinds of findings that could be returned to them, it is clear that they have a durable preference not to know genetic information about themselves.¹⁹⁹ When assessing the benefit and harms of this CMN, the question is whether “the potentially significant harms of patients or participants misreporting their preferences on a consent form and forgoing valuable health information outweighs the harms of not respecting the preferences of a handful of strong refusers who do not opt-out.”²⁰⁰

In this case, economic harms could result from employers or insurers making discriminatory decisions— such as the loss of a job or an inability to acquire affordable health insurance—on the basis of an individual’s genetic status.²⁰¹ Psychosocial harms might include things like depression, anxiety, and stigmatization.²⁰² Though there has long been concern about harms associated with genetic testing,²⁰³ these arguments have largely been based on hypothetical concerns and have not been supported by emerging evidence.²⁰⁴ A comprehensive review of the literature on psychosocial harms associated with genetic testing concluded that current evidence suggests that:

the original ELSI concerns were unfounded, exaggerated, or, at a minimum, misdirected. At least in the contexts that have been most studied, large negative impacts have not been found in the vast majority of people studied.²⁰⁵

A similar story can be told about economic harms, where the Genetic Information Nondiscrimination Act (GINA)²⁰⁶ has served to mitigate most of the concerns about genetic discrimination. Even in other contexts not covered by GINA (such as long-term care and life-insurance), there remains little evidence of widespread discrimination.²⁰⁷

In summary, it appears that there are strong potential direct benefits to the “weak refusers” and somewhat weaker benefits to their relatives. Though there are hypothetical risks to “strong refusers”, there is scant evidence in the literature to support these concerns.²⁰⁸ On net, we argue that the benefits of a CMN in the context of the RNTK generally outweigh the harms. The value of providing some participants with potentially life-saving information is very high, and the burden of depriving a “strong refuser” of an explicit choice is low given that it is possible for them to still opt-out if they feel strongly enough to independently raise the issue with the researchers.

2. Harms to Trust and its Tangible Consequences

A more concerning harm potentially associated with CMNs is that nudging could be viewed as duplicitous, and nudgees’ perceptions of being manipulated might erode their trust in medical researchers.²⁰⁹ Harm to trust is concerning in and of itself but could also have significant tangible consequences when lack of trust in medical research develops into a general lack of trust in medical institutions. The line between medical research and medical care is often blurry, especially because some forms of medical care are only available in research settings, such as experimental drugs for diseases with no known treatment.²¹⁰ It can be difficult to distinguish between being treated as a medical research participant and being treated as a patient. As a result, loss of trust in medical researchers may translate to loss in trust in medical institutions more broadly. If CMNs are perceived as duplicitous by undermining nudgees’ trust in medical institutions writ large, and if that causes those nudgees to be more hesitant in seeking care and less likely to heed advice from their doctors, it could indirectly contribute to worse health outcomes. This concern could arise for any individual, but it may be particularly salient for members of groups who, because of historical mistreatment, may have less trust and confidence in medical care and research to begin with.²¹¹

Historical evidence suggests that lack of trust in medical establishments is associated with significantly worse medical outcomes and with health disparities.²¹² Some medical mistrust is linked to historical medical racism.²¹³ In addition to past examples of egregious medical misconduct, such as the Tuskegee Syphilis Study,²¹⁴ everyday experiences of medical racism are thought to contribute to racialized medical mistrust.²¹⁵ Disparities in care and personal experiences of discrimination as well as the general mistrust of social institutions also contribute to medical mistrust.²¹⁶ CMNs could exacerbate problems of medical mistrust if they are interpreted as a way for researchers to disrespectfully avoid communication with research participants. To counteract such negative effects, institutions must first take actions to prove themselves worthy of medical trust. Beyond that, efforts to deliberately build trust and increase transparency have proven essential to recruiting populations with a history of research abuse to participate in subsequent medical research.²¹⁷ A nudge that appears to mask choice could be perceived as deceptive, and thus potentially in tension with trust-building recommendations.

Worries about medical mistrust are heightened when we have good reason to believe that the CMN concerns a decision that provokes a variety of opinions. In such cases, nudged endorsement of any option is likely to raise suspicions. For example, there is extensive literature on public opinion about sharing newborn blood spots and genomic data. While it is difficult to draw absolute conclusions about public opinion, we do know that people’s views on the broad sharing of genomic data are diverse and that there is at least a significant minority of people who have strong views about the acceptability of broad data and sample sharing.²¹⁸ Data on views about NBS research is similar.²¹⁹ Given the range of perspectives on NBS research and GDS, it is reasonable to exercise caution about endorsing a CMN model.²²⁰ In fact, recognizing the uncertainty in the public polling data, the GDS policy explicitly prohibits an opt-out approach (although as discussed above, the policy does seem to leave room for some strategies that look CMN adjacent).²²¹ Despite a similarly diverse range of views on the sharing of bloodspots, state-level NBS policies still largely include a CMN approach.²²²

Refraining from CMNs in an effort to improve transparency and build trust might appear to be in tension with the purpose of implementing CMNs to receive the associated benefits. Potential nudgers face a challenge: they must implement CMNs in an effective, yet also transparent and respectful way. The degree to which relevant information must be provided and questions invited will depend on the sensitivity of the choice being masked and the historical levels of distrust among research participant groups—the greater the likelihood of the nudge causing distrust, the greater the need for transparency.²²³ Consider members of the Havasupai Tribe. Researchers violated the trust of a group of tribe members—who initially consented to participate in diabetes research study that was expected to generate research benefits—by using their stored blood samples were used without their consent for genetic testing that discredited cultural creation myths in a way that was deeply offensive to many of the Havasupai.²²⁴ Nudging members of the Havasupai Tribe to give blood samples could warrant greater transparency and communication efforts than would be required in nudging the general population of new parents—who have not had such negative experiences with research misconduct—to participate in NBS. Though more transparent nudges might be less effective, the tradeoff would be justified in the aforementioned case. The extent to which perceived duplicity from choice-masking is likely to have harmful consequences depends on past relationships between nudgers and nudgees, as well as the sensitivity of the content of the nudge.

C. Unfair Distribution of CMN Burdens

Another important consideration in evaluating CMNs is whether their burdens are distributed unfairly. CMNs are most likely to impose a burden when participants have difficulty opting out of a default option they would prefer not to receive.²²⁵ This unfair distribution of resistibility may occur at the individual or the group level. On the individual level, it would be unfair to impose an extreme burden on a research participant solely for the purpose of lessening a benefit to others.²²⁶ Nudges can be considered unfairly distributed on the group level when the individual characteristics that make resisting nudges difficult fall along group lines or are related to group membership.²²⁷ Unfair distribution of nudging effects along group lines is especially concerning because it has the potential to exacerbate underlying social injustices, particularly if the distribution disadvantages already disadvantaged groups. If it is harder for some nudgees to resist a nudge than others, it would be especially unfair if those who had the most difficulty resisting the nudge also belong to a disadvantaged group in society.²²⁸

There are two reasons we might worry about unfairly distributing nudge burdens along group lines. First, nudges could concentrate burdens or harms in certain groups without providing benefits to members of the same group. This might be the case, for example, if a CMN—today—nudged members of the Havasupai Tribe to donate genetic samples to be used in research unlikely to benefit the donors.²²⁹ Given historical conflicts, such a nudge could do serious damage to group trust in medical institutions. Second, CMNs might be less resistible to some groups than to others. Even if opt-out mechanisms are easy and readily available, the nature of the CMN requires that participants self-identify as choosing to opt-out of it.²³⁰ The option is only available to those who actively select it, which involves articulating one’s preferences, seemingly contradicting the research team’s preferences.²³¹ Members of groups with cultural values that emphasize deference to authority, groups with low literacy, or members of groups that are generally socially vulnerable may feel less confident in articulating their concerns or desire to activate the opt-out process.

At the same time, it is important to ask why individuals might want to resist a given nudge. Members of groups who have difficulty resisting a nudge may also have more to lose from resisting. For example, in a recent paper, Mrkva and colleagues suggest that well-implemented nudges can be used to reduce choice disparities.²³² Through a series of studies, they show that individuals with low-socioeconomic status were more impacted by nudges than those with high-socioeconomic status, and that individuals with lower levels of domain knowledge and numeracy were more impacted by nudges than individuals with high levels of domain knowledge and numeracy.²³³ These findings suggest that carefully constructed nudges that benefit low-socioeconomic status individuals help to reduce inequalities and promote equity. If the nudge is legitimate and beneficial overall, it might be more unfair not to nudge than it would be to nudge. This consideration might depend, to some extent, on whether the nudge is paternalistic or justified by appealing to society, generally. Means-paternalistic nudges are designed to offset potential disproportionate burdens with benefits to the nudgees on their own terms.²³⁴ Thus, it is more important that nudges justified by their contribution to the social good are shown to not be unfair than it is for means-paternalistic nudges that promote a nudgee’s own interests.²³⁵ At any rate, nudgers must be especially aware that individuals belonging to vulnerable social groups may be less likely to self-identify as having non-default preferences and should, as a result, avoid designing nudges that would take advantage of that dynamic.

D. Lack of Ethically Superior Feasible Alternatives

Instances of paternalism—even justified and resistible ones, as any nudge that satisfies the three considerations listed above will be—should be taken seriously since they involve acting on behalf of other people. Nudges are intended to be effective and should therefore, at least to some degree, restrict the nudgee’s scope of self-determination. To justify the use of a CMN, the researcher must show that there are not feasible effective alternatives to achieve a nudge’s given aim.²³⁶ While comparison with feasible alternatives must ultimately be done on a case-by-case basis, cognitive biases are strong indicators of a lack of feasible alternatives.

1. Counteracting Cognitive Biases

Nudges will often be the best feasible way to achieve a desired outcome when they address decisions in which choosers are likely to be affected by cognitive biases because cognitive biases often prevent people from rationally pursuing what they would prefer in light of their considered beliefs, attitudes, and preferences. In a decision that requires careful deliberation, someone might display cognitive bias by becoming frustrated and choosing rashly without making an effort to consider all available options; nudges can counteract that effect. Not all kinds of choices are equally subject to cognitive biases. Typically, the greater the likelihood of a decision inducing cognitive bias, the more justified a nudge will be.²³⁷ The emotionally charged nature of medical decision-making is likely to contribute to cognitive bias.²³⁸ A decision-maker’s emotions often affects her choice, particularly if the choice will have effects far into the future.²³⁹ Research has shown, for example, that when choosers feel stressed by the decision they are asked to make, they tend to unwittingly overestimate how long they will feel that way and decide differently than they would in a calmer state.²⁴⁰ What is at stake in a particular decision also has implications for how much cognitive bias is likely to enter into the choice. It has been suggested that in very low stakes decisions, choosers might experience low processing motivation and consequently select the most convenient option without careful consideration.²⁴¹ At the same time, very high stakes decisions might also induce cognitive bias because they can be overwhelming and make reasonable deliberation very difficult.²⁴² Nudges are more likely to outperform feasible alternatives in highly emotionally charged situations and when decision stakes are particularly high or low.²⁴³

Returning to our cases, an excellent example of cognitive biases can be found in the RNTK debate. There is extensive psychological literature about individuals’ poor ability to predict how future negative events will impact our emotional wellbeing.²⁴⁴ This affective forecasting literature suggests that while people think that future unfortunate events will be devastating, most humans are actually more emotionally resilient than we realize and we have the ability to adapt to even terrible news.²⁴⁵ This is particularly true in the context of genetic testing, where most reactions to unfavorable information are mild and transient.²⁴⁶ This kind of widespread cognitive bias might be used to support implementation of a CMN. In contrast, consider the cognitive biases inherent in the way that consent is obtained for NBS. The typical procedure is to ask new parents, in the days after giving birth, to review a stack of paperwork, including the notification that they can opt out of allowing their child’s bloodspot to be used for future research.²⁴⁷ This chaotic time might not be an ideal moment to implement a CMN because the parents’ ability to cognitively engage and deliberate might be compromised.

2. Least Restrictive Alternative

Stakes and emotional valence can, at best, give a rough indication of whether a CMN is likely to be better than feasible alternatives, however. To make that determination in specific cases, it is necessary to directly compare alternatives in terms of their relative harms and benefits. This comparison can be done with the widely used less restrictive alternatives (LRA) test, which compares a proposal to alternatives along two dimensions: restrictiveness and effectiveness.²⁴⁸ In the case of CMNs, restrictiveness measures how harmful the nudge would likely be and effectiveness measures how likely it is to achieve its aims.²⁴⁹ The test is easy to use when one alternative is dominant—meaning that it is both more effective and less restrictive. When no alternative under comparison is superior on both axes—such as when the nudge is more effective but slightly more restrictive—there is no shortcut to careful deliberation and weighing of alternatives. Restrictiveness should only be tolerated if the effectiveness is far greater, but harms to autonomy and social justice put a limit on the level of restrictiveness that can be tolerated.

To illustrate this kind of analysis, consider different alternatives to a CMN in the context of the RNTK. As one alternative, investigators could include more detail in the consent form about the kinds of secondary findings that might be discovered, thinking that this could increase understanding. Such a proposition would have to be tested, however, because there is extensive literature that calls into question participants’ ability to engage with and internalize basic concepts being conveyed in consent forms.²⁵⁰ Another alternative might be to schedule a dedicated conversation with a genetics counselor about the RNTK. However, this would be prohibitively resource intensive and might not even be feasible given the well-documented shortage of genetic counselors.²⁵¹ A final option would be to recontact participants who refused secondary findings to give them another chance to make a choice. This option, however, risks alerting those who really did not want to know about the existence of a positive secondary finding. Given these options, it seems reasonable to conclude that a CMN is the best option, assuming that there is a mechanism for self-identified refusers to easily opt-out.

An analysis of LRAs becomes more complicated in situations where there is only societal benefit—like the NBS and GDS cases—involving a two-step assessment of how much scientific value is lost when you switch from a CMN to an explicit opt-out mechanism. This will first typically require an empirical analysis of the difference in size and composition of the research resource under a CMN and other consent strategies. For instance, how many fewer people (and which ones) will not agree to give their sample for future research if explicitly asked? It then becomes important to ask whether that lost value significantly undermines the research goals. For example, if researchers want to create a genomic database to study common conditions, losing a small portion of a repository might not be deleterious unless those losses came from a group of particular interest. A research resource being used to study rare diseases or variants might have more scientific justification to maximize the size and representativeness of the repository.

VII. Conclusion

Effective use of CMNs in medical research has the potential to benefit individuals in ways they themselves endorse, to help people collectively, and to promote important social objectives.²⁵² For the kind of purely informational nudges we are concerned with here, the benefits extend beyond the individuals being nudged; our basis for justifying CMNs thus extends beyond individuals to consider collective and societal benefit. As our examples of the RNTK genetic information about oneself, NBS, and GDS all show, these benefits are not merely hypothetical.²⁵³ But neither are their costs. While CMNs have great potential positive effects, they could also transgress individuals’ autonomy interests. If medical researchers are to use CMNs, it is important that they do so in legally and ethically acceptable ways. Thus, the challenge facing researchers is to design nudges that are likely to be effective overall, but that are also easily resistible to those who wish to resist.²⁵⁴ Where the sweet spot is when soliciting questions without raising concerns will depend on the context of a particular nudge. The primary aim of this paper has been to provide a framework for designing acceptable CMNs that capture the potential benefits without incurring the potential costs.

Our analysis of the reasonable person standard and U.S. federal research regulations shows how CMNs fit into current legal regulations.²⁵⁵ Research regulations require medical researchers to provide participants with enough information so that participants understand the research they intend to participate in, but do not require explicit choice.²⁵⁶ Importantly, they refer to the reasonable person standard to determine what counts as sufficient.²⁵⁷ The structure of the RPS helps us see how reasonableness is evaluated by directing us to consider all relevant considerations and weigh them in a reasonable way.²⁵⁸ For the medical research questions that we are interested in, a reasonable person would weigh her interests in having a full range of choices explicitly presented against how both she and others are likely to be affected by the nudge. CMNs that respect protected autonomy domains and conform to the considerations discussed in Part VI of this article are expected to be acceptable to reasonable persons.²⁵⁹

We expect legally and ethically acceptable CMNs to have legitimate policy aims, generate more benefits than harms, not unfairly distribute burdens, and lack feasible alternatives. The stringency of our acceptability considerations, however, will vary with the context. In some cases, historical facts might make a certain choice particularly sensitive and render masking inappropriate,²⁶⁰ Moreover, significantly more masking may be acceptable when benefits are likely to accrue directly to the individual being nudged. There is precedent for this kind of sliding scale of acceptability.²⁶¹ Our acceptability considerations for CMNs are similarly responsive to the social and historical context of the choice, the directness of the (positive and negative) consequences likely to accrue to research participants, and the degree of ethical controversy surrounding the nudge. Return once more to the RNTK example. While a nudge toward receiving medically actionable secondary findings would likely be appropriate in most contexts, it would not be in all contexts.²⁶²

The most likely challenges to CMNs come from critics who claim that CMNs are likely to violate fundamental autonomy interests. While we acknowledge that CMNs may not be welcome on every view of autonomy, they are compatible with the most important features of autonomy; nudged research participants can still be the driving forces of their own lives. The easy resistibility built in to acceptable CMNs undermines objections that CMNs would force research participants into unwillingly and unwittingly accepting researchers’ default decisions. Medical research is complicated and can be difficult for participants to understand; thoughtfully designed CMNs have an important role to play in gently helping large numbers of research participants reach decision outcomes that really are best for them and their communities.